The neural representation of time Richard B Ivry
and Rebecca MC Spencer This review summarizes recent investigations of temporal processing. We focus on motor and perceptual tasks in which crucial events span hundreds of milliseconds. One key question concerns whether the representation of temporal information is dependent on a specialized system, distributed across a network of neural regions, or computed in a local task-dependent manner. Consistent with the specialized system framework, the cerebellum is associated with various tasks that require precise timing. Computational models of timing mechanisms within the cerebellar cortex are beginning to motivate physiological studies. Emphasis has also been placed on the basal ganglia as a specialized timing system, particularly for longer intervals. We outline an alternative hypothesis in which this structure is associated with decision processes.

In this review we focus on tasks in the shorter range. Even within this range, the phrase ‘temporal processing’ may refer to very different phenomena. Temporal order tasks require an ordinal judgment, indicating the order of successive stimulus events. These types of judgments are affected by the rate of temporal integration. Other tasks require a metrical judgment that involves the analysis of elapsed time. The assessment of duration might be either explicit, as in a duration discrimination task, or implicit, as in eyeblink conditioning, in which the response must be precisely timed to be adaptive. We restrict our review here to tasks in which timing would appear to be metrical.

Addresses 3210 Tolman Hall, Department of Psychology, University of California, Berkeley, California 94720-1650 USA
e-mail: [email protected]

Does the existence of temporal regularities imply that some process is dedicated to representing time? Temporal regularities could be explicitly represented, reflecting a dedicated internal timing mechanism. Dedicated timing could be performed locally, or result from the operation of a specialized neural structure or distributed network (Figure 1). Alternatively, temporal regularities could be an emergent property, reflecting the fact that dynamic processes such as those involved in coordinating limbs for action [4,5] or selective attending in perception [6] occur in time.

Current Opinion in Neurobiology 2004, 14:225–232 This review comes from a themed issue on Cognitive neuroscience Edited by John Gabrieli and Elisabeth A Murray 0959-4388/$ – see front matter ß 2004 Elsevier Ltd. All rights reserved. DOI 10.1016/j.conb.2004.03.013

Is there a specialized neural region for millisecond timing?

Lesion studies

The representation of temporal information remains one of the most elusive concepts for neurobiology. Unlike vision and audition, there are no dedicated sensors for time. Yet the passage of time is as perceptually salient as the color of an apple or the timbre of a tuba.

Various lines of evidence indicate that the cerebellar cortex provides a precise representation of the temporal relationship between successive events. Perhaps the most compelling evidence comes from studies of eyeblink conditioning in which the conditioned response (CR) must be timed to occur just before the unconditioned stimulus. Studies consistently demonstrate that the CR is disrupted following lesions of the cerebellum [7]. Whereas associative mechanisms operate at various levels within the cerebellum [8], accurate timing of the CR is dependent on the cerebellar cortex [9,10]. Knockout species lacking the capability for long-term depression (LTD) at the parallel fiber–Purkinje cell synapses fail to exhibit adaptive timing [11].

Fraisse [1] was the first to emphasize that a discontinuity in our sense of time was evident around 2–3 s. Lewis and Miall [2,3] argue that timing in the shorter range is ‘automatic’, reflecting the engagement of processes associated with the production of skilled movements. Longer range timing is hypothesized to be ‘cognitive’, dependent on neural systems associated with attention and working memory.

The movements of patients with cerebellar lesions are characterized by a breakdown of the timing between muscular events. For example, these patients are inaccurate in throwing, in part because of increased variability in timing the opening of the hand with respect to arm rotation [12,13]. However, such deficits do not necessarily imply the involvement of an explicit timing signal. Hand opening might be triggered by cerebellar computations of

Abbreviations CR conditioned response fMRI functional magnetic resonance imaging PD Parkinson’s disease SMA supplementary motor area

Introduction: scope of the review

www.sciencedirect.com

Current Opinion in Neurobiology 2004, 14:225–232

226 Cognitive neuroscience

Figure 1

(a)

Specialized timing

(b)

(c)

Distributed network timing

Localized timing

Current Opinion in Neurobiology

General frameworks of the neural mechanisms for timing. (a) The specialized timing model is based on the idea that a particular neural region is uniquely capable of representing temporal information and that this system is recruited when this form of processing is required. This example illustrates the cerebellum as a specialized system. (b) In the distributed network timing model, the representation of temporal information results from the interactions within a set of neural structures. (c) The local timing model does not entail a dedicated timing system. Rather, temporal information is computed within the neural structures required for a particular task.

the dynamic transitions required between successive states, a form of forward modeling by the cerebellum [14]. The manner in which a task is conceptualized can influence how timing is achieved [15,16]. Patients with cerebellar lesions show increased variability on temporal production tasks, such as rhythmic tapping, or during the production of isolated movements with a specified target duration [17]. However, these patients are unimpaired when the periodic movements are smooth and continuous. This dissociation is consistent with the hypothesis that tasks involving discontinuities or salient features embody an event structure. The cerebellum provides the signals specifying the timing of these events, similar to the way in which the conditioned and unconditioned stimuli in eyeblink conditioning define two salient events. By contrast, continuous movements lack this event structure and temporal regularities are an emergent property reflecting the operation of another control parameter (e.g. angular velocity) [15,16,17]. Current Opinion in Neurobiology 2004, 14:225–232

Harrington et al. [18] failed to observe consistent increased temporal variability on production or perception tasks in patients with unilateral cerebellar lesions. However, a subset of patients with lesions encompassing the superior cerebellum exhibited increased variability on the production task and a marginally significant increase on the perception task. Interestingly, disruption of eyeblink conditioning is also more pronounced in patients with superior cerebellar lesions when compared to those with inferior lesions [7]. Lesion studies have also implicated the basal ganglia in temporal processing. This work, conducted within the framework of the influential scalar timing model [19], has generally involved intervals up to 40 s. Timing within this range is assumed to involve a set of separable components including a pacemaker, accumulator, gating mechanism, and decision processes, in which the output of the accumulator is compared to reference memory of stored intervals. The basal ganglia are hypothesized to be a www.sciencedirect.com

The neural representation of time Ivry and Spencer 227

crucial component of the pacemaker/accumulator process. In contrast to normal animals, rats with striatal lesions fail to increase the rate of lever pressing at the time of an expected reward [20]. Additionally, dopaminergic agents lead to a systematic distortion of timed responses: agonists and antagonists lead to a shortening and lengthening, respectively, of perceived time. These results are consistent with the hypothesis that dopamine levels affect the speed of an internal pacemaker. However, dopamine deficiencies could disrupt memory functions [21] or, as will be discussed in the conclusions, alter decision processes. Whether or not the basal ganglia are involved in timing in the range of hundreds of milliseconds remains unclear. Some studies report time perception deficits in patients with Parkinson’s disease (PD) ([22,23] but see [24]), and pharmacological manipulations in normal individuals can alter temporal acuity [25]. Graeber et al. [26] report that a subset of PD patients show a marked bias on a speech perception task in which the discrimination between two consonants is temporally cued. The patients’ judgments suggested that the crucial interval was underestimated, consistent with the idea that dopamine depletion leads to the slowing of an internal pacemaker. Time production studies in the milliseconds range, however, are inconsistent with this hypothesis. PD patients tend to speed up on finger tapping tasks [18,27,28]. Moreover, the literature is inconsistent in terms of whether or not PD patients show increased temporal variability on production tasks [18,27–29]. PD can be problematic for studying basal ganglia dysfunction given the widespread reduction in dopamine. A forthcoming study uses an alternative approach, testing patients with chronic focal lesions of the striatum [30]. Surprisingly, these patients exhibited no impairment on a finger tapping task. Although lesion studies of timing in the milliseconds range have focused on the cerebellum and basal ganglia, a cortical locus cannot be dismissed. Various lines of evidence suggest that temporal processing could be differentially affected by lesions of the right and left hemispheres [31], or that the hemispheres integrate information at different speeds [32–34]. Surprisingly, few studies have tested patients with cortical lesions on time perception and production tasks. In one such study, patients with right hemisphere lesions were impaired on a duration discrimination task for intervals of 300 and 600 ms [35]. The impairment was attributed to attentional processes required for gating timing signals into working memory. Similarly, repetitive transcranial magnetic stimulation (TMS) over right prefrontal cortex in neurologically healthy individuals altered the perception of intervals spanning 5–15 s [36]. www.sciencedirect.com

Neuroimaging studies

In contrast to the relatively sparse lesion literature, the number of neuroimaging studies of temporal processing has increased exponentially in recent years. Two recent reviews have summarized this work [3,37]. Given this, our review focuses on four new functional magnetic resonance imaging (fMRI) papers involving duration discrimination tasks with intervals in the millisecond range [2,38–40]. Lewis and Miall [2] asked participants to judge the duration of horizontal length of a visual stimulus. In the duration conditions, the stimuli could vary around 0.6 s or 3 s. Compared to the length conditions, duration judgments were associated with increased activation in prefrontal, insula, premotor (lateral and supplementary motor area [SMA]), and parietal cortices. Moreover, activation specific to the 0.6 s condition was observed in the right temporal lobe and left cerebellar hemisphere. Activation specific to the 3 s condition was observed in left parietal cortex and posterior cingulate. A similar pattern was found in a study using intervals around 1 s [39]. Compared to a temporal order judgment control task, duration discrimination led to increased activation in right prefrontal cortex, SMA, and left cerebellum. Basal ganglia activation during the duration tasks was not found in either study. However, two fMRI studies have reported putamen activation during duration discrimination tasks. In one study [38], the stimulus duration was centered around 700 ms and participants judged either duration or brightness. Cortical foci in the duration task included bilateral prefrontal, temporal, and inferior parietal cortices, as well as the SMA, the left premotor area, and the right insula. Basal ganglia activation was restricted to the left putamen. Cerebellar activation in the vermis was similar in both tasks, suggesting that this region was not specifically recruited for temporal processing. A similar cortical network was observed active in a duration discrimination task with auditory stimuli when performance was compared to that during rest [40]. However, essentially the same areas were also recruited in the frequency discrimination control task. Right putamen activation was greater for the duration task, but only in a restricted analysis that used a liberal statistical threshold to evaluate activation within this region. Cerebellar coverage was limited in this study and thus a similar analysis could not be performed. Studies of time production have focused on tasks in which rhythmic complexity is varied [41–43]. Identification of time-specific areas in such studies is difficult as baseline conditions also require the production of timed movements. An alternative approach is to look at changes in brain activation when participants learn movement patterns in which the sequence of finger responses is fixed, the sequence of inter-response intervals is fixed, or both Current Opinion in Neurobiology 2004, 14:225–232

228 Cognitive neuroscience

[44]. The inferior temporal gyrus and the lateral cerebellum were the only activation foci specific to temporal learning.

Physiological analysis of temporal processing The literature is replete with sophisticated computational models for the representation of temporal information. Delay line mechanisms, operating in the microsecond range, have been proposed to underlie sound localization. Differences in the time required for neural signals to traverse fixed distances, coupled with coincidence detectors, can be exploited in simple networks to determine the horizontal position of a sound source [45]. Given the speed of neural conduction times, such mechanisms are unlikely to produce sufficient intervals for timing in the hundreds of milliseconds range [46], and would certainly fail for longer intervals. For temporal phenomena over longer ranges, physiological mechanisms fall into two broad classes [20,47]. One class is based on the idea that temporal codes are formed through the operation of oscillatory processes. As noted, the scalar timing model posits that the representation of duration entails a clockcounter mechanism [48]. Although this model was developed for tasks spanning many seconds, researchers have assumed that similar mechanisms operate at short intervals. The other class can be defined by models in which the continuum of time can be represented without oscillatory events; these are termed ‘spectral models’. Spectral models posit the translation of a temporal code into a spatial code. Different intervals are represented by the activation of non-overlapping neural elements, perhaps because of delays introduced by the stochastic properties of slow physiological processes [47,49]. This does not mean that such delay properties are fixed; learning mechanisms could be used to shape input and output relationships. Alternatively, the dynamics of time-varying physiological events might be used to represent and produce temporal information [47,49,50]. Physiological studies have just begun to test these models. Leon and Shadlen [51] recorded from neurons in inferior parietal cortex of the monkey while the animals judged the duration of visual events centered around 300 ms or 800 ms. Psychometric functions derived from neural ensembles approximated the animals’ behavior, suggesting that these cells provide a representation of time. Consistent with this idea, physiological mechanisms such as slow inhibitory post-synaptic potentials (IPSPs) are ubiquitous in the nervous system, and could serve as the building block for temporal processing [49,52]. According to this view, timing information is locally computed in a task-dependent manner [51,53,54]. Alternatively, the activity of these parietal neurons could reflect decision processes given that similar brain-behaCurrent Opinion in Neurobiology 2004, 14:225–232

vior relationships are observed for a variety of psychophysical tasks [55,56]. According to this view, the stimulus duration might be computed upstream (e.g. in the cerebellum) then transmitted to neurons associated with specific response systems (e.g. eye movements as in the study by Leon and Shadlen [51]). Evoked potential studies in humans are also consistent with the hypothesis that cortical signals indicate the evolution of decision processes [37,57–59]. Neurophysiological studies of eyeblink conditioning have provided the most detailed analysis of the emergence of time-dependent behavior [10,11]. As noted earlier, CRs persist after lesions of the cerebellar cortex, but the adaptive timing is abolished [9,11]. Various models of the cerebellar cortex have been proposed, instantiating different forms of spectral coding [60,61]. In one model, interactions between granule cells and Golgi cells produce a range of delays for the efficacy of parallel fiber input to Purkinje cells [62]. A representation of the unconditioned stimulus conveyed by climbing fibers is used to strengthen those inputs that are tuned to drive the CR at the optimal time.

Conclusions The recent neuroimaging literature is consistent with the hypothesis that the cerebellum is engaged during tasks requiring the precise representation of temporal information. This includes motor sequence learning [44], rhythmic tapping [41–43], duration discrimination [2,39], phoneme perception [63], and attentional anticipation [64]. Whereas imaging studies are best viewed in terms of a sufficiency argument, lesion studies provide a stronger test of necessity [65]. Again, the data from human and animal studies indicate that lesions of the cerebellum are associated with increased temporal variability. We do not wish to suggest that the instantiation of temporal processing within the cerebellum is generic; rather we assume that subregions within the cerebellar cortex will be recruited for timing in a task-dependent manner [66]. Thus, we emphasize a general computational principle of the cerebellum. Neuroanatomical considerations make it unlikely that internal timing would be task independent; such a hypothesis would be exceedingly complex in terms of the mapping between inputs and outputs across diverse tasks. The current evidence does not preclude distributed models or hypotheses that assign a central role for timing to another specialized system, such as the basal ganglia. As reviewed here, the results of imaging and lesion studies are ambiguous with respect to the role of the basal ganglia in timing short intervals. A clear dissociation between the cerebellar and the basal ganglia contributions on temporal processing tasks www.sciencedirect.com

The neural representation of time Ivry and Spencer 229

Figure 2

(a)

(b) Motor cortex

Prefrontal cortex

Basal ganglia

Limbic system

Oculomotor cortex Current Opinion in Neurobiology

Hypothesized gating operation of the basal ganglia as part of a decision making process. (a) Potentiated cortical representations provide input to the basal ganglia. The output from the basal ganglia reflects selected representations that have reached threshold. (From Gazzaniga et al. [71], art work by F Forney.) (b) The functional consequences of this gating process will depend on input–output circuitry [72]. For example, the motor loop will trigger overt movements, whereas the prefrontal loop involves the updating of working memory.

remains elusive, primarily because similar deficits have been observed in patients with lesions of either structure ([22,28,67,68] but see [68]). The cerebellar hypothesis offers a parsimonious account over a broad set of tasks, and neurobiologically feasible models have been developed. Nonetheless, a specialized system hypothesis must be able to account for similar patterns of performance following damage to distinct systems. As a starting point, we propose that the basal ganglia are an integral part of decision processes, operating as a threshold mechanism (Figure 2). Activations into the basal ganglia are gated such that only those reaching threshold are implemented [69]. The activation functions for different decisions can reflect multiple factors, such as goals, sensory inputs, and contextual information. These representations engage in a competitive process for control. According to this view, the basal ganglia ensure that response implementation or working memory updating does not occur until a criterion level of activation is reached. Dopamine inputs to the striatum modulate threshold settings, providing one mechanism by which the competition can be biased. Thresholds for reinforced actions are lowered, increasing the likelihood of implementation, even if the input patterns are unchanged. Although this hypothesis is intended to describe the role of the basal ganglia in response or set selection, it provides a novel perspective of impairments on temporal processing tasks associated with basal ganglia dysfunction. Consider the perception of intervals on the order of multiple seconds. Judging the amount of elapsed time for such intervals is attention mediated [70], or what has been called cognitive timing [2,3]. One way such timing could be achieved is by monitoring the number of updates of working memory, a form of an accumulator model. www.sciencedirect.com

Dopamine levels distort the perception of time (Figure 3a). In the threshold model, dopamine agonists would lower thresholds, leading to more frequent updates and a criterion number of updates would be reached earlier. Likewise, time perception would be lengthened when thresholds are raised by dopamine antagonists. This hypothesis can also be applied to short intervals without postulating a direct role for the basal ganglia in the representation of time. Dopamine agents would again be expected to distort perceived time [25]. Moreover, an appealing feature of this hypothesis is that the same mechanism can account for PD akinesia, the difficulty to initiate movement. In the absence of dopamine, thresholds are elevated. The gating operation would thus be delayed, requiring extended accrual for a particular activation pattern. This simple model would not account for PD patients’ impairments in judging the duration of a short stimulus, given our assumption that the representation of stimulus duration is derived in the cerebellum. However, it is reasonable to assume that the depletion of dopamine not only changes the threshold setting but also introduces additional noise into these settings. In this manner, perceptual judgments would be more variable, reflecting threshold fluctuations or response biases. However, such deficits should not be specific to duration discrimination, a prediction not supported by one study [22]. With one additional modification, the threshold model can account for the tendency of PD patients to speed up during repetitive movements [22,27,28], a result that seems at odds with pacemaker models. We assume that dopamine primarily acts as a long-term modulator of thresholds; over the short term, thresholds will be sensitive to recent Current Opinion in Neurobiology 2004, 14:225–232

230 Cognitive neuroscience

Figure 3

(a)

(b)

Normal DAag

Activation

Activation

DAant

Time

Time

Current Opinion in Neurobiology

Gating of activated representations through threshold adjustment. The green line represents the activation signal that serves as an input to the basal ganglia. Drop-lines indicate time of gating for a particular threshold setting. (a) Dopamine agonists lower the threshold, leading to the gating operation being invoked with less activation. Dopamine antagonists raise the threshold. This mechanism can be applied to understand the effects of dopamine depletion in Parkinson’s disease (PD) or the effects of dopamine-based reinforcement. For the latter, reinforcement signals serve to lower thresholds, leading to increased probability of an input reaching threshold in the future. (b) Tendency of PD patients to speed up during unpaced finger tapping could result from short-term modulation of elevated thresholds. After each output, the system resets and a new activation signal accrues for the next response. The gaps indicate that the input to the gating mechanism might not be immediate, but builds up near the target time, reflecting activation in upstream systems that determine onset time (e.g. cerebellum). Assuming that variation in the activation function is random, gating will tend to occur earlier as the threshold is reduced over cycles.

context effects (Figure 3b). Thus, a threshold recently triggered will be lowered, especially when the initial state is inflated. As a result, successive cycles through a circuit will gradually decrease in cycle rate, even if the input remains constant. We recognize that one could reinterpret cerebellar timing deficits within a non-timing hypothesis, similar to what we have attempted with respect to the basal ganglia. Our intent here is to offer functional hypotheses that can motivate new empirical and computational endeavors. Such efforts will be necessary as part of the continuing efforts to disentangle the contributions of different neural systems to temporal processing.

resting baseline. Third, scanning parameters ensured excellent coverage of cortex and cerebellum. 3.

Lewis PA, Miall RC: Distinct systems for automatic and cognitively controlled time measurement: evidence from neuroimaging. Curr Opin Neurobiol 2003, 13:250-255.

4.

Semjen A: Emergent versus programmed temporal properties of movement sequences. In Time and Mind Edited by Helfrich H. Seattle, WA: Hogrefe & Huber; 1996:25-43.

5.

Turvey MT: Preliminaries to a theory of action with reference to vision. In Perceiving, Acting and Knowing. Edited by Bransford R. Hillsdale NJ: Lawerence Erlbaum; 1977:211-265.

6.

McAuley JD, Jones MR: Modeling effects of rhythmic context on perceived duration: a comparison of interval and entrainment approaches to short-interval timing. J Exp Psychol Hum Percept Perform 2003, 29:1102-1125.

7.

Gerwig M, Dimitrova A, Kolb FP, Maschke M, Brol B, Kunnel A, Boring D, Thilmann AF, Forsting M, Diener HC et al.: Comparison of eyeblink conditioning in patients with superior and posterior inferior cerebellar lesions. Brain 2003, 126:71-94.

8.

Bao SW, Chen L, Kim JJ, Thompson RF: Cerebellar cortical inhibition and classical eyeblink conditioning. Proc Natl Acad Sci U S A 2002, 99:1592-1597.

9.

Perrett SP, Ruiz BP, Mauk MD: Cerebellar cortex lesions disrupt learning-dependent timing of conditioned eyelid responses. J Neurosci 1993, 13:1708-1718.

Acknowledgements We are grateful to T Verstynen, J Diedrichsen, S Ell, and S Keele for their comments. This work was supported by National Institutes of Health Grants NS30256, NS17778, NS33504, and NS40813.

References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as:  of special interest  of outstanding interest 1.

Fraisse P: The Psychology of Time. New York: Harper & Row; 1963.

2. 

Lewis PA, Miall RC: Brain activation patterns during measurement of sub- and supra-second intervals. Neuropsychologia 2003, 41:1583-1592. The authors present a study in which fMRI was used to identify brain systems engaged during a duration discrimination task. The control task required judgments of stimulus length. Three aspects of this study are noteworthy. First, the psychophysical methods ensured that experimental and control tasks were of comparable difficulty and, unlike many similar studies, participants had to attend to the stimulus for its entire duration in both conditions. Second, primary analyses focus on direct comparison of the two conditions rather than a comparison of each to a

Current Opinion in Neurobiology 2004, 14:225–232

10. Kotani S, Kawahara S, Kirino Y: Purkinje cell activity during  learning a new timing in classical eyeblink conditioning. Brain Res 2003, 994:193-202. The authors made single unit recordings in the cerebellar lobule HVI of decerebrate guinea pigs during eyeblink conditioning with the initial interval between the conditioned and unconditioned stimuli set to 250 ms, followed by transfer to an interval of 400 ms. The majority of cells showed time-dependent response to the conditioned stimulus, although the pattern could be excitatory, inhibitory, or complex. However, the profile of individual cells was uncorrelated for the two inter-response intervals (ISIs). Their results stand against the hypothesis that the new timing is learned by simple shift of output properties of learning-related cells; rather, learning of new timing required recruitment of different neural populations within the cerebellum. 11. Koekkoek SKE, Hulscher HC, Dortland BR, Hensbroek RA,  Elgersma Y, Ruigrok TJH, DeZeeuw CI: Cerebellar LTD and www.sciencedirect.com

The neural representation of time Ivry and Spencer 231

learning-dependent timing of conditioned eyelid responses. Science 2003, 301:1736-1739. The authors conducted eyeblink conditioning in transgenic L7-PKCi mice, a knockout species in which LTD at the parallel fiber–Purkinje synapse in cerebellar cortex is impaired by inhibition of protein kinase C. These animals developed conditioned responses over a multi-day training session, although at lower levels compared to wild type littermates. The peak amplitude was much earlier in the mutants and remained maladaptively timed when the interval between the conditioned and the unconditioned stimulus was lengthened. 12. Timmann D, Watts S, Hore J: Failure of cerebellar patients to time finger opening precisely causes ball high-low inaccuracy in overarm throws. J Neurophysiol 1999, 2:103-114. 13. McNaughton S, Timmann D, Watts S, Hore J: Overarm throwing speed in cerebellar subjects: effect of timing of ball release. Exp Brain Res 2004, 154:470-478. 14. Kawato M: Internal models for motor control and trajectory planning. Curr Opin Neurobiol 1999, 9:718-727. 15. Ivry RB, Diedrichsen J, Spencer RMC, Hazeltine E, Semjen A: A cognitive neuroscience perspective on bimanual coordination and interference. In Interlimb Coordination. Edited by Swinnen SP, Duyens J. Norwell, MA: Kluwer Academic; in press.

24. Ivry RB, Keele SW, Diener HC: Dissociation of the lateral and medial cerebellum in movement timing and movement execution. Exp Brain Res 1988, 73:167-180. 25. Rammsayer T: Neuropharmacological evidence for different timing mechanisms in humans. Q J Exp Psychol 1999, 52:273-286. 26. Graeber S, Hertrich I, Daum I, Spieker S, Ackermann H: Speech perception deficits in Parkinson’s disease: underestimation of time intervals compromises identification of durational phonetic contrasts. Brain Lang 2002, 82:65-74. 27. Elsinger CL, Rao SM, Zimbelman JL, Reynolds NC, Blindauer KA, Hoffman RG: Neural basis for impaired time reproduction in Parkinson’s disease: An fMRI study. J Int Neuropsychol Soc 2003, 9:1088-1098. 28. Ivry RB, Keele SW: Timing functions of the cerebellum. J Cogn Neurosci 1989, 1:136-152. 29. O’Boyle DJ, Freeman JS, Cody FWJ: The accuracy and precision of timing of self-paced, repetitive movements in subjects with Parkinson’s disease. Brain 1996, 119:51-70. 30. Aparicio PD, Diedrichsen J, Ivry RB: Effects of focal basal ganglia lesions on timing and force control. Brain Cogn in press.

16. Ivry RB, Spencer RMC, Zelaznik HN, Diedrichsen J: The cerebellum and event timing. Ann N Y Acad Sci 2002, 978:302-317.

31. Kagerer FA, Wittmann M, Szelag E, Von Steinbuchel N: Cortical involvement in temporal reproduction: evidence for differential roles of the hemispheres. Neuropsychologia 2002, 40:357-366.

17. Spencer RMC, Zelaznik HN, Diedrichsen J, Ivry RB: Disrupted  timing of discontinuous movements by cerebellar lesions. Science 2003, 300:1437-1439. Previous studies (Zelaznik et al. [73]) had shown that individual differences in timing acuity were not correlated between rhythmic production tasks in which the movements involved discontinuities between successive cycles (slight pause or contact with response surface) and rhythmic production tasks in which the movements were produced in a smooth, continuous manner. As predicted, patients with cerebellar lesions were selectively impaired on the tasks involving discontinuities. The lack of impairment on the continuous tasks is especially striking given that these movements required coordination across multiple joints and involved significant interaction torques.

32. Fitch RH, Miller S, Tallal P: Neurobiology of speech perception. Annu Rev Neurosci 1997, 20:331-353.

18. Harrington DL, Lee RR, Boyd LA, Rapcsak SZ, Knight RT: Does the  representation of time depend on the cerebellum? Effect of cerebellar stroke. Brain 2004, 127:561-574. The authors provide the most thorough attempt to date to specify subregions within the cerebellum that are essential for accurate timing on production and perception tasks. The authors focus on the lack of a consistent impairment in the group of patients, arguing that the results challenge a central role for the cerebellum in temporal processing. However, this acceptance of the null hypothesis is questionable given that patients with lesions in the superior cerebellum exhibited increased variability on a rhythmic tapping task and a marginally significant increase (p ¼ 0.07) on a duration discrimination task. The superior cerebellum was defined as lobules superior to lobulus semilunaris inferior (above Crus II), and the lesions extended into the dentate nucleus in some of the patients.

36. Koch G, Oliveri M, Torriero S, Caltagirone C: Underestimation of time perception after repetitive transcranial magnetic stimulation. Neurology 2003, 60:1844-1846.

19. Meck WH, Benson AM: Dissecting the brain’s internal clock:  how frontal-striatal circuitry keeps time and shifts attention. Brain Cogn 2002, 48:195-211. The authors present a concise review that focuses on hypothesized functions of frontal–striatal circuitry for temporal processing tasks. The article provides an overview of the component operations that constitute the scalar timing model, including timing, attention, and memory processes, and reviews how lesion and imaging studies suggest differential roles for cortical and basal ganglia regions in the instantiation of these processes.

33. Robertson LC, Ivry R: Hemispheric asymmetries: attention to visual and auditory primitives. Curr Dir Psychol Sci 2000, 9:59-63. 34. Poeppel D: The analysis of speech in different temporal integration windows: cerebral lateralization as ‘asymmetric sampling time’. Speech Commun 2003, 41:245-255. 35. Harrington DL, Haaland KY, Knight RT: Cortical networks underlying mechanisms of time perception. J Neurosci 1998, 18:1085-1095.

37. Macar F, Lejeune H, Bonnet M, Ferrara A, Pouthas V, Vidal F, Maquet P: Activation of the supplementary motor area and of attentional networks during temporal processing. Exp Brain Res 2002, 142:475-485. 38. Ferrandez AM, Hugueville L, Lehericy S, Poline JB, Marsault C, Pouthas V: Basal ganglia and supplementary motor area subtend duration and perception: an fMRI study. Neuroimage 2003, 19:1532-1544. 39. Smith A, Taylor E, Lidzba K, Rubia K: A right hemispheric frontocerebellar network for time discrimination of several hundreds of milliseconds. Neuroimage 2003, 20:344-350. 40. Nenadic I, Gaser C, Volz H-P, Rammsayer T, Hager F, Sauer H: Processing of temporal information in the basal ganglia: new evidence from fMRI. Exp Brain Res 2003, 148:238-246. 41. Mayville JM, Jantzen KJ, Fuchs A, Steinberg FL, Kelso JAS: Cortical and subcortical networks underlying syncopated and synchronized coordination revealed using fMRI. Hum Brain Mapp 2002, 17:214-229.

20. Matell MS, Meck WH, Nicolelis MAL: Interval timing and the encoding of signal duration by ensembles of cortical and striatal neurons. Behav Neurosci 2003, 117:760-773.

42. Ullen F, Forssberg H, Ehrsson HH: Neural networks for the coordination of the hands in time. J Neurophysiol 2003, 89:1126-1135.

21. Malapani C, Deweer B, Gibbon J: Separating storage from retrieval dysfunction of temporal memory in Parkinson’s disease. J Cogn Neurosci 2002, 14:311-322.

43. Dhamala M, Pagnoni G, Wiesenfeld K, Zink CF, Martin M, Berns GS: Neural correlates of the complexity of rhythmic finger tapping. Neuroimage 2003, 20:918-926.

22. Harrington DL, Haaland KY, Hermanowicz N: Temporal processing in the basal ganglia. Neuropsychology 1998, 12:3-12.

44. Sakai K, Ramnani N, Passingham RE: Learning of sequences of  finger movements and timing: frontal lobe and action-oriented representation. J Neurophysiol 2002, 88:2035-2046. The authors report an fMRI experiment of motor learning in which two variables were manipulated. First, the sequence of finger movements was either specified or random. Second, the timing of the stimuli specifying

23. Artieda J, Pastor MA, Lacruz F, Obeso JA: Temporal discrimination is abnormal in Parkinson’s disease. Brain 1992, 115:199-210. www.sciencedirect.com

Current Opinion in Neurobiology 2004, 14:225–232

232 Cognitive neuroscience

the responses was fixed or random. fMRI analyses focused on activation increases and decreases over training sessions associated with the finger sequence alone, the timing sequence alone, or both. Areas activated in the two conditions in which temporal learning was possible were limited to the contralateral inferior parietal lobe and ipsilateral cerebellar hemisphere. 45. Konishi M: Coding in auditory space. Annu Rev Neurosci 2003, 26:31-55. 46. Braitenberg V: Is the cerebellar cortex a biological clock in the millisecond range? Prog Brain Res 1967, 25:334-346. 47. Buonomano DV, Karmarkar UR: How do we tell time? Neuroscientist 2002, 8:42-51. 48. Gibbon J: Scalar expectancy theory and Weber’s Law in animal timing. Psychol Rev 1977, 84:279-325. 49. Buonomano DV: Timing of neural responses in cortical  organotypic slices. Proc Natl Acad Sci U S A 2003, 100:4897-4902. Reliably timed action potentials occurring up to 300 ms after stimulus onset are observed in an organotypic cortical slice preparation obtained from rat auditory cortex. The late responses are attributed to the intrinsic dynamics of the neural network, consistent with modeling work in which slow physiological processes are hypothesized to be capable of processing temporal information in the order of hundreds of milliseconds. The authors hypothesize that timing in this range could be a local process, reflected in the dynamics of neurons recruited in a task-specific manner. 50. Hopfield JJ, Brody CD: What is a moment? Transient synchrony as a collective mechanism for spatiotemporal integration. Proc Natl Acad Sci U S A 2001, 98:1282-1287. 51. Leon MI, Shadlen MN: Representation of time by neurons in  the posterior parietal cortex of the macaque. Neuron 2003, 38:317-327. Animals were trained to judge the duration of a visual stimulus and indicate their response with an eye movement after an extended delay. Neural activity was time-dependent during the stimulus period, indicating that the cells were not simply recruited at the time of the response. However, it remains plausible that the neural activity reflects the dynamics of decision processes with temporal coding originating upstream. 52. Buonomano DV: Decoding temporal information: a model based on short-term synaptic plasticity. J Neurosci 2000, 20:1129-1141. 53. Brody CD, Hernandez A, Antonio Z, Romo R: Timing and neural encoding of somatosensory parametric working memory in macaque prefrontal cortex. Cereb Cortex 2003, 13:1196-1207. 54. Roux S, Coulmance M, Riehle A: Context-related representation of timing processes in monkey motor cortex. Eur J Neurosci 2003, 18:1011-1016. 55. Shadlen MN, Newsome WT: Neural basis of a perceptual decision in the parietal cortex (area LIP) of the rhesus monkey. J Neurophysiol 2001, 86:1916-1936. 56. Sawamura H, Shima K, Tanji J: Numerical representation for action in the parietal cortex of the monkey. Nature 2002, 415:918-922. 57. Pfeuty M, Ragot R, Pouthas V: When time is up: CNV time course differentiates the roles of the hemispheres in the discrimination of short tone durations. Exp Brain Res 2003, 151:373-379. 58. Pfeuty M, Ragot R, Pouthas V: Processes involved in tempo perception: a CNV analysis. Psychophysiology 2003, 40:69-76. 59. Monfort V, Pouthas V: Effects of working memory demands on frontal slow waves in time-interval reproduction tasks in humans. Neurosci Lett 2003, 343:195-199. 60. Ohyama T, Nores WL, Murphy M, Mauk MD: What the cerebellum  computes. Trends Neurosci 2003, 26:222-227.

Current Opinion in Neurobiology 2004, 14:225–232

The authors present a computationally oriented review of how eyeblink conditioning provides a model system for understanding feedforward control mechanisms used to support sensorimotor learning. The review emphasizes the importance of error signals in this form of learning, how error signals are used to support associative processes that operate at various levels and how they can shape behavior that is temporally specific. 61. Fiala JC, Grossberg S, Bullock D: Metabotropic glutamate receptor activation in cerebellar Purkinje cells as substrate for adaptive timing of the classically conditioned eye-blink response. J Neurosci 1996, 16:3760-3774. 62. Buonomano DV, Mauk MD: Neural network model of the cerebellum: temporal discrimination and the timing of motor responses. Neural Comput 1994, 6:38-55. 63. Mathiak K, Hertrich I, Grodd W, Ackermann H: Cerebellum and  speech perception: a functional magnetic resonance imaging study. J Cogn Neurosci 2002, 14:902-912. The authors used fMRI to study the neural correlates for processing the phonemic cue of closure duration for stop consonants. Acoustically, this cue can be signaled temporally by the duration of silence or by a combination of spectral and temporal information. Importantly, people might be unaware of this manipulation; thus, the two imaging conditions are essentially identical. Nonetheless, when the cues were purely temporal, activation was restricted to left frontal and right cerebellar foci; by contrast, the inclusion of spectral cues resulted in a shift of activation to left superior temporal lobe. 64. Dreher J-C, Grafman J: The roles of the cerebellum and basal  ganglia in timing and error prediction. Eur J Neurosci 2002, 16:1609-1619. Both the cerebellum and the basal ganglia have been implicated in similar cognitive operations including timing, error prediction, and the coordination of attentional set. In an attempt to directly evaluate these hypotheses, the authors conducted an fMRI study of task switching in which the timing and order of changes in a task set were manipulated. Cerebellar activation was sensitive to the temporal manipulation. By contrast, striatal activation was sensitive to task order predictability, interpreted as reflecting a role in error prediction. Activation in neither cerebellum nor striatum increased during attention switching compared to a control condition with similar working memory demands. 65. Price CJ, Friston KJ: Degeneracy and cognitive anatomy. Trends Cogn Sci 2002, 6:416-421. 66. Ivry RB: The representation of temporal information in perception and motor control. Curr Opin Neurobiol 1996, 6:851-857. 67. Spencer RMC, Ivry RB: Comparison of patients with Parkinson’s disease or cerebellar lesions in the production of periodic movements involving event-based or emergent timing. Brain Cogn in press. 68. Diedrichsen J, Ivry R, Pressing J: Cerebellar and basal ganglia contributions to interval timing. In Functional and Neural Mechanisms of Interval Timing. Edited by Meck WH: CRC Press; 2003:457-483. 69. Berns GS, Sejnowski TJ: How the basal ganglia make decisions. In The Neurobiology of Decision Making. Edited by Damasio A, Damasio H, Christen Y: Springer-Verlag; 1996:101-113. 70. Block RA, Zakay D: Prospective and retrospective duration judgments: a meta-analytic review. Psychon Bull Rev 1997, 4:184-197. 71. Gazzaniga M, Ivry R, Mangun GR: Cognitive neuroscience: the biology of the mind edn 2. New York, NY: W.W. Norton and Company, Inc; 2002. 72. Alexander GE, Crutcher MD: Functional architecture of basal ganglia circuits: neural substrates of parallel processing. Trends Neurosci 1990, 13:266-271. 73. Zelaznik HN, Spencer RMC, Ivry RB: Dissociation of explicit and implicit timing in repetitive tapping and drawing movements, J Exp Psychol Hum Percept Perform 28:575-588.

www.sciencedirect.com